AbouttheEditors
MatthiasDehmer studiedmathematicsattheUniversityofSiegen,Siegen, GermanyandearnedhisPhDincomputersciencefromtheTechnical UniversityofDarmstadt,Darmstadt,Germany.Afterward,hewasaresearch fellowatViennaBioCenter,Austria,ViennaUniversityofTechnology,and UniversityofCoimbra,Portugal.Heobtainedhishabilitationinapplied discretemathematicsfromtheViennaUniversityofTechnology.Currently,he isaprofessoratUMIT—TheHealthandLifeSciencesUniversity,Austria.His researchinterestsareindatascience,BigData,complexnetworks,machine learning,andinformationtheory.Hehaspublishedmorethan220publicationsinappliedmathematics,computerscience,datascience,andrelated disciplines.
FrankEmmert-Streib studiedphysicsattheUniversityofSiegen,Germany, andearnedhisPhDintheoreticalphysicsfromtheUniversityofBremen, Bremen,Germany.HewasapostdoctoralfellowintheUnitedStatesbefore becomingafacultymemberattheCenterforCancerResearchattheQueen’s UniversityBelfast,UK.Currently,heisaprofessorintheDepartmentof SignalProcessingatTampereUniversityofTechnology,Finland.Hisresearch interestsareinthefieldofcomputationalbiology,datascienceandanalytics inthedevelopmentandapplicationofmethodsfromstatistics,andmachine learningfortheanalysisofBigDatafromgenomics,finance,andbusiness.
Contributors
RamozaAhsan WorcesterPolytechnicUniversity Worcester,Massachusetts
BetsyBarry EmoryUniversity Atlanta,Georgia
AmparoAlonso-Betanzos UniversidadedaCoru˜na ACoru˜na,Spain
NancyH.Brinson UniversityofTexasatAustin Austin,Texas
Ver´onicaBol´on-Canedo UniversidadedaCoru˜na ACoru˜na,Spain
Yi-TingChen NationalChiaoTungUniversity Hsinchu,Taiwan
MatthewS.Eastin UniversityofTexasatAustin Austin,Texas
AlfredFr¨uh Universit¨atZ¨urich Z¨urich,Switzerland
StefanA.Kaiser
IndependentResearcher Wassenberg,Germany
CaitlinKuhlman WorcesterPolytechnicUniversity Worcester,Massachusetts
MonikaLaner FernfachhochschuleSchweiz Brig,Switzerland
AlessandroMantelero PolytechnicUniversityofTurin Turin,Italy
NoeliaS´anchez-Maro˜no UniversidadedaCoru˜na ACoru˜na,Spain
RodicaNeamtu WorcesterPolytechnicUniversity Worcester,Massachusetts
BeatricePaoli FernfachhochschuleSchweiz Brig,Switzerland
WolfgangPietsch TechnischeUniversit¨atM¨unchen Munich,Germany
ElkeRundensteiner WorcesterPolytechnicUniversity Worcester,Massachusetts
x Contributors
Beth-AnneSchuelke-Leech UniversityofWindsor Windsor,Ontario,Canada
JouriSemenov FernfachhochschuleSchweiz Brig,Switzerland
EdwardW.Sun KEDGEBusinessSchool Talence,France
FlorentThouvenin Universit¨atZ¨urich Z¨urich,Switzerland
BeatT¨odtli FernfachhochschuleSchweiz Brig,Switzerland
GiuseppeVaciago UniversityofInsubria Varese,Italy
BartvanderSloot TilburgUniversity Tilburg,theNetherlands
ZhongWang BeijingAcademyofSocialSciences Beijing,China
XiaohuaWang ChineseAcademyofSciences Beijing,China
RolfH.Weber Universit¨atZ¨urich Z¨urich,Switzerland
ofindividualswithregardtotheprocessingofpersonaldataina
∗ AlessandroMantelero,PolytechnicUniversityofTurin,istheauthorofsections “Introduction:Thelegalchallengesoftheuseofdata”and“Useofdatafordecision-making purposes:Fromindividualtocollectivedimensionofdataprocessing.”GiuseppeVaciago, UniversityofInsubria,istheauthorofsection“Dataprediction:Socialcontrolandsocial surveillance.”
Introduction:Thelegalchallengesoftheuseofdata
TherearemanydefinitionsofBigData,whichdifferdependingonthe specificdiscipline.Mostofthedefinitionsfocusonthegrowingtechnological abilitytocollect,process,andextractnewandpredictiveknowledgefroma bulkofdatacharacterizedbyagreatvolume,velocity,andvariety.∗
However,intermsofprotectionofindividualrights,themainissuesdo notonlyconcernthevolume,velocity,andvarietyofprocesseddata,butalso theanalysisofdata,usingsoftwaretoextractnewandpredictiveknowledge fordecision-makingpurposes.Therefore,inthiscontribution,thedefinition ofBigDataencompassesbothBigDataandBigDataanalytics.†
TheadventofBigDatahassuggestedanewparadigminsocialempiricalstudies,inwhichthetraditionalapproachadoptedinstatisticalstudiesis complementedorreplacedbyBigDataanalysis.Thisnewparadigmischaracterizedbytherelevantroleplayedbydatavisualization,whichmakesit possibletheanalysisofreal-timedatastreamstogettheirtrajectoryand predictfuturetrendspossible[3].Moreover,largeamountsofdatamakeit possibletouseunsupervisedmachine-learningalgorithmstodiscoverhidden correlationsbetweenvariablesthatcharacterizelargedatasets.
Thiskindofapproach,whichisbasedontheemergingcorrelationsamong data,leadssocialinvestigationtoadoptanewstrategy,inwhichthereare nopreexistingresearchhypothesestobeverifiedthroughempiricalstatisticalstudies.BigDataanalyticssuggestpossiblecorrelations,whichconstitute perse theresearchhypothesis:datashowthepotentialrelationsbetweenfacts orbehavior.Nevertheless,theserelationsarenotgroundedoncausationand, forthisreason,shouldbefurtherinvestigatedusingthetraditionalstatistical method.
Assumingthatdatatrendssuggestcorrelationsandconsequentresearch hypotheses,atthemomentofdatacollectiononlyverygeneralresearch hypothesesarepossible,asthepotentialdatapatternsarestillunknown. Therefore,thespecificpurposeofdataprocessingcanbeidentifiedonlyat alatertime,whencorrelationsrevealtheusefulnessofsomeinformationto detectspecificaspects.Onlyatthattime,thegivenpurposeoftheuseof informationbecomesevident,alsowithregardtofurtheranalysesconducted withtraditionalstatisticalmethods[4].
∗ Theterm“BigData”usuallyidentifiesextremelylargedatasetsthatmaybeanalyzed computationallytoextractinferencesaboutdatapatterns,trends,andcorrelations.AccordingtotheInternationalTelecommunicationUnion,BigDataare“aparadigmforenabling thecollection,storage,management,analysis,andvisualization,potentiallyunderreal-time constraints,ofextensivedatasetswithheterogeneouscharacteristics”[1].
† Thistermisusedtoidentifycomputationaltechnologiesthatanalyzelargeamountsof datatouncoverhiddenpatterns,trends,andcorrelations.AccordingtotheEuropeanUnion AgencyforNetworkandInformationSecurity,thetermBigDataanalytics“referstothe wholedatamanagementlifecycleofcollecting,organizing,andanalysingdatatodiscover patterns,toinfersituationsorstates,topredictandtounderstandbehaviors”[2].
Ontheotherhand,therearealgorithms,suchassupervisedmachinelearningalgorithms,thatneedapreliminarytrainingphase.Inthisstage,a supervisorusesdatatrainingsetstocorrecttheerrorsofthemachine,orienting thealgorithmtowardcorrectassociations.Inthissense,supervisedmachinelearningalgorithmsrequireapriordefinitionofthepurposeoftheuseof data,identifyingthegoalthatthemachineshouldreachthroughautonomous processingofallavailabledata.
Inthiscase,althoughthepurposeofdatauseisdefinedinthetraining phase,themannerinwhichdataareprocessedandthefinaloutcomeofdata miningremainlargelyunknown.Infact,thesealgorithmsareblackboxesand theirinternaldynamicsarepartiallyunpredictable.∗
Bothdatavisualizationandmachine-learningapplicationsposerelevant questionsintermsofBigDataprocessing,whichwillbeaddressedinthe followingsections.Howisitpossibletodefinethespecificpurposeofdata processingatthemomentofdatacollection,whenthecorrelationssuggested byanalyticsareunknownatthattime?Ifdifferentsourcesofdataareused inmachinetrainingandrunninglearningalgorithms,howcandatasubjects knowthespecificpurposeoftheuseoftheirinformationingivenmachinelearningapplications?
Thesequestionsclearlyshowthetensionthatcharacterizestheapplication ofthetraditionaldataprotectionprinciplesintheBigDatacontext.Butthis isnottheonlycrucialaspect:theverynotionofpersonaldataisbecoming moreundefined.RunningBigDataanalyticsoverlargedatasetscouldmake itdifficulttodistinguishbetweenpersonaldataandanonymousdata,aswell asbetweensensitivedataandnonsensitivedata.
Variousstudieshavedemonstratedhowinformationstoredinanonymized datasetscanbepartiallyreidentified,insomecaseswithoutexpensivetechnicalsolutions[5–12].Thissuggestsgoingbeyondthetraditionaldichotomy betweenpersonalandanonymousdataandrepresentingthisdistinctionas ascalethatmovesfrompersonalidentifiedinformationtoaggregateddata. Betweentheseextremes,thelevelofanonymizationisproportionaltothe effort,intermsoftime,resourcesandcosts,whichisrequiredtoreidentify information.
Finally,withregardtosensitivedata,BigDataanalyticsmakeitpossibletousenonsensitivedatatoinfersensitiveinformation,suchasinformationconcerningreligiouspracticesextractedfromlocationdataandmobility patterns[13].
Againstthisbackground,theexistingdataprotectionregulationsandthe ongoingproposals[14,15]remainlargelyfocusedonthetraditionalmainpillarsoftheso-calledfourthgenerationofdataprotectionlaws[16]:the notice
∗ See,e.g.,ZhangM.,“GooglePhotosTagsTwoAfrican-AmericansAsGorillas ThroughFacialRecognitionSoftware,” Forbes,July1,2015.http://www.forbes.com/sites/ mzhang/2015/07/01/google-photos-tags-two-african-americans-as-gorillas-through-facialrecognition-software/#36b529227b63(accessedMarch23,2016).
andconsent model(i.e.,aninformed,freelygiven,andspecificconsent) [17–21],∗ thepurposelimitationprinciple[24,25],andtheminimization principle.
Forthisreason,thefollowingsectionsinvestigatethelimitsandcriticisms oftheexistinglegalframeworkandthepossibleoptionstoprovideadequate answerstothenewchallengesofBigDataprocessing.Inthislight,thischapter isdividedintothreemainsections.
Thefirstsectionfocusesonthetraditionalparadigmofdataprotection andontheprovisions,primarilyinthenewEUGeneralDataProtection Regulation(Regulation(EU)2016/679,hereafterGDPR),thatcanbeused tosafeguardindividualrightsinBigDataprocessing.
Thesecondsectiongoesbeyondtheexistinglegalframeworkand,inthe lightofthepathopenedbytheguidelinesonBigDataadoptedbytheCouncilofEurope,suggestsabroaderapproachthatencompassesthecollective dimensionofdataprotection.ThisdimensionoftencharacterizesBigData applicationsandleadstoassesstheethicalandsocialimpactsofdatauses, whichassumeanimportantroleinmanyBigDatacontexts.
ThelastsectiondealswiththeuseofBigDatatoanticipatefrauddetection andtopreventcrime.Inthislight,thenewDirective(EU)2016/680† isbriefly analyzed.
Datacollectionanddataprocessing:Thefundamentalsofdata protectionregulations
Beforeconsideringthedifferentreasonsthatinducethelawtoprotect personalinformation,itshouldbenotedthatEuropeanlegalsystemsdonot recognizethesamebroadnotionoftherighttoprivacythatexistsinU.S. jurisprudence.‡ Atthesametime,intheEuropeancountries,dataprotection lawsdonotdrawtheiroriginsfromtheEuropeanideaofprivacyandits relatedcaselaw.
∗ SeeArticles6and7,Regulation(EU)2016/679oftheEuropeanParliamentandofthe CouncilofApril27,2016ontheprotectionofnaturalpersonswithregardtotheprocessing ofpersonaldataandonthefreemovementofsuchdata,andrepealingDirective95/46/EC (GeneralDataProtectionRegulation).Differently,intheUnitedStates,thetraditional approachbasedonvarioussectorialregulationshasunderestimatedtheroleplayedbyuser’s choice,adoptingamarket-orientedstrategy.Nevertheless,theguidelinesadoptedbythe U.S.administrationsin2012[14]seemtosuggestadifferentapproach,reinforcingselfdetermination[8,22,23].
† Directive(EU)2016/680ontheprotectionofnaturalpersonswithregardtothe processingofpersonaldatabycompetentauthoritiesforthepurposesoftheprevention, investigation,detectionorprosecutionofcriminaloffencesortheexecutionofcriminalpenalties,andonthefreemovementofsuchdata,andrepealingCouncilFrameworkDecision 2008/977/JHA.
‡ Withregardtothenotionofrighttoprivacy(andinbrief),intheUnitedStatesthe righttoprivacycoversabroadareathatgoesfrominformationalprivacytotherightof self-determinationinprivatelifedecisions.Ontheotherhand,inEuropeancountries,this rightmainlyfocusesonthefirstaspectandisrelatedtomediaactivities[26–31].
Europeandataprotectionregulations,sincetheiroriginsinthesecond halfofthelastcentury,focusedoninformationregardingindividuals,without distinguishingbetweenpublicorprivateinformation[32].Comparedwiththe righttoprivacy,theissuesregardingtheprotectionofpersonaldatahavebeen morerecentlyrecognizedbylaw,bothintheUnitedStatesandEurope[33]. Thisdatesfromthe1960s,whereastheprimitiveeraoftherighttoprivacy wasattheendofthenineteenthcentury,whenthepennypressassumed asignificantroleinlimitingtheprivacyofthepeoplebelongingtoupper classes[34].
Inthelightoftheabove,theanalysisofthefundamentalsofdataprocessingshouldstartfromtheeffectsofthecomputerrevolutionthathappenedin thelate1950s.Theadventofcomputersanditssocialimpactledtothefirst regulationsondataprotectionandposedthefirstpillarsofthearchitecture ofthepresentlegalframework.
Thefirstgenerationsofdataprotectionregulationswerecharacterizedbya nationalapproach.Theywereadoptedindifferenttimesbynationallegislators andweredifferentwithregardtotheextensionofthesafeguardsprovidedand theremediesoffered.
Thenotionofdataprotectionwasoriginallybasedontheideaofcontrol overinformation,asconfirmedbytheliteratureofthatperiod[35–37].The migrationfromdustypaperarchivestocomputermemorieswasaCopernicanrevolutionwhich,forthefirsttimeinhistory,permittedtheaggregation ofinformationabouteverycitizenthatwaspreviouslyspreadoverdifferent archives[38].
Thefirstdataprotectionregulationsweretheanswertotherisingconcern ofcitizensaboutsocialcontrol,asthenewbigmainframecomputersgave governments[16,38–41]andlargecorporationstheopportunitytocollectand managelargeamountofpersonalinformation[16,42].Inthissense,thelegal systemsgaveindividualstheopportunitytohaveasortofcountercontrolover thecollecteddata[16,38,43].
Thepurposeoftheregulationswasnottospreadanddemocratizepower overinformationbuttoincreasetheleveloftransparencyaboutdataprocessingandsafeguardtherighttoaccesstoinformation.Citizensfeltthey weremonitored,andthelawgavethemtheopportunitytoknowwhocontrolledtheirdata,whichkindofinformationwascollected,andforwhich purposes.
Themandatorynotificationsofnewdatabases,registration,licensingprocedures,andindependentauthorities[16,44]werethefundamentalelements ofthesenewregulations.Theywerenecessarytoknowwhohadcontrolover informationandtomonitordataprocessing.Anotherkeycomponentwasthe righttoaccess,whichallowscitizenstoaskdataownersaboutthewayin whichinformationisusedand,consequently,abouttheexerciseoftheirpower overinformation.Finally,theentirepicturewascompletedbythecreationof adhoc publicauthoritiestosafeguardandenforcecitizen’srights,exercise controloverdataowners,andreactagainstabuses.
Inthismodel,therewasnospaceforindividualconsent,duetotheeconomiccontextofthatperiod.Thecollectionofinformationwasmainlymade bypublicentitiesforpurposesrelatedtopublicinterests,wasmandatory, andtherewasnospaceofautonomyintermsofnegotiationaboutpersonal information.Atthesametime,personalinformationdidnothaveaneconomicvalueforprivatecompanies:dataaboutclientsandsupplierswere mainlyusedforoperationalfunctionsregardingtheexecutionofcompany activities.
Anotherelementthatcontributedtoexcludetheroleofself-determination wasthelackofknowledge,theextremedifficultyforordinarypeopletounderstandtheuse,andthemodeofoperationofmainframes.ThecomputermainframeswereasortofmodernGod,withsacralattendants,aselectednumber oftechnicianswhowereabletousethisnewequipment.Inthisscenario,it didnotmakesensetogivecitizensthechancetochoose,astheywereunable tounderstandthewayinwhichtheirdatawereprocessed.
Inconclusion,duringthe1970sandthefirstpartofthe1980softhe lastcentury,legislatorslaidthefoundationsfordataprotectionregulationsin manyEuropeancountriesandoutsideEurope,asaresultofthetechnological andsocialchangesofthatperiod.Thesefirstregulationsdefinedtheinitial coreofdataprotection(i.e.,transparency,rightstoaccess,anddataprotection authorities),whichisstillpresentintheexistinglegalframework.
TheEuropeanUnionmodel:FromtheDataProtection DirectivetotheGeneralDataProtectionRegulation
Theperiodfromthemid-1980stothe1990swascharacterizednotonlyby therisingofauniformapproachtodataprotectionregulationamongthemembersoftheEuropeanUnion,butalsobyachangeintheregulatoryparadigm, duetothenewtechnological,social,andeconomicscenarios.
Homecomputersenteredthemarketinthelate1970stobecomecommon duringthe1980s.Thiswastheneweraofdistributedcomputers,inwhicha lotofpeopleboughtapersonalcomputertocollectandprocessinformation.
Thebigmainframecomputersbecamethesmalldesktoppersonalcomputers,witharelativelylowcost.Consequently,thecomputationalcapacity wasnolongeranexclusiveprivilegeofgovernmentsandbigcompaniesbut becameaccessibletomanyentitiesandconsumers.
Thisperiodwitnessedanothertransformationinvolvingdirectmarketing, whichwasnolongerbasedontheconceptofmailorderandmovedtoward computerizeddirectmarketingsolutions ∗ Thenewformsofmarketingwere basedoncustomerprofilingandrequiredextensivedatacollectiontoapply
∗ Althoughdirectmarketinghasitsrootsinmailorderservices,whichwerebasedon personalizedletter(e.g.,usingthenameandsurnameofaddressees)andgeneralgroup profiling(e.g.,usingcensusinformationtogroupaddresseesinsocialandeconomicclasses), theuseofcomputerequipmentincreasedthelevelofmanipulationofconsumerinformation andgenerateddetailedconsumer’sprofiles[45,46].
dataminingsoftware.Themainpurposeofprofilingwastosuggestasuitable commercialproposaltoanyconsumer.
Thiswasaninnovativeapplicationofdataprocessingdrivenbynewpurposes.Informationwasnolongercollectedtosupportsupplychains,logistics, andorders,buttosellthebestproducttoeachuser.Asaresult,thedata subjectbecamethefocusoftheprocess,andpersonalinformationacquired aneconomicandbusinessvalue,givenitsroleinsales.
Thesechangesinthetechnologicalandbusinessframeworkscreatednew requestsfromsocietytolegislators,ascitizenswantedtohavethechanceto negotiatetheirpersonaldataandgainsomethinginreturn.
AlthoughthenewgenerationsoftheEuropeandataprotectionlawsplaced personalinformationwithinthecontextoffundamentalrights,∗ themaingoal oftheseregulationswastopursueeconomicinterestsrelatedtothefreeflow ofpersonaldata.ThisisalsoaffirmedbytheDirective95/46/EC,† which representsboththegeneralframeworkandthesynthesisofthissecondwave ofdataprotectionlaws.‡
However,therootsofdataprotectionremainedinthecontextofpersonalityrights.Therefore,theEuropeanapproachislessmarket-orientedthanit happensinotherlegalsystems.Thedirectivealsorecognizesthefundamental roleofpublicauthoritiesinprotectingdatasubjectsagainstunwilledorunfair exploitationoftheirpersonalinformationformarketingpurposes.
Boththetheoreticalmodeloffundamentalrights,basedonselfdetermination,andtherisingdata-driveneconomyhighlightedtheimportance ofuserconsentinconsumerdataprocessing.Consentdoesnotonlyrepresent anexpressionofchoicewithregardtotheuseofpersonalityrightsbythird partiesbutisalsoaninstrumenttonegotiatetheeconomicvalueofpersonal information.
Inthisnewdata-driveneconomy,personaldatacannotbeexploitedfor businesspurposeswithoutanyinvolvementofdatasubjects.Itisnecessary thatindividualsbecomepartofthenegotiation,asdataarenolongerused mainlybygovernmentagenciesforpublicpurposesbutalsobyprivatecompanieswithmonetaryrevenues[49,50].
∗ SeeCouncilofEurope,ConventionfortheProtectionofIndividualswithregard toAutomaticProcessingofPersonalData,openedforsignatureonJanuary28,1981 andenteredintoforceonOctober1,1985.http://conventions.coe.int/Treaty/Commun/ QueVoulezVous.asp?NT=108&CL=ENG(accessedFebruary27,2014);OECD,Annex totheRecommendationoftheCouncilof23rdSeptember1980:GuidelinesontheProtectionofPrivacyandTransborderFlowsofPersonalData.http://www.oecd.org/internet/ ieconomy/oecdguidelinesontheprotectionofprivacyandtransborderflowsofpersonaldata.htm# preface(accessedFebruary27,2014).
† Directive95/46/ECoftheEuropeanParliamentandoftheCouncilof24October1995 ontheprotectionofindividualswithregardtotheprocessingofpersonaldataandonthe freemovementofsuchdata[1995]OJL281/31.
‡ TheEUDirective95/46/EChasadualnature,asitwaswrittenonthebasisofthe existingnationaldataprotectionlaws,inordertoharmonizethem,butatthesametimeit alsoprovidedanewsetofrules.SeetherecitalsinthepreambletotheDirective95/46/EC [47,48].
Effectiveself-determinationindataprocessing,bothintermsofprotectionandeconomicexploitationofpersonalityrights,cannotbeobtained withoutadequateandpriornotice.∗ Forthisreason,the noticeandconsent model† addedanewlayertotheexistingparadigmbasedontransparencyand access[17].
Finally,itisimportanttohighlightthat,duringthe1980sand1990s,data analysisincreasedinquality,butitslevelofcomplexitywasstilllimited.Consequently,consumerswereabletounderstandthegeneralcorrelationbetween datacollectionandrelatedpurposesofdataprocessing(e.g.,profilingusers, offeringcustomizedservices,orgoods).Atthattime,informedconsentand self-determinationwerelargelyconsideredassynonyms,butthisischanged now,intheBigDataera.
TheadventofBigDataanalyticshascreatedadifferenteconomicand technologicalscenario,withdirectconsequencesontheadequacyofthelegal frameworkadoptedtosafeguardpersonalinformation.Thenewenvironment ismainlydigitalandcharacterizedbyanincreasingconcentrationofinformationinthehandsofafewentities,bothpublicandprivate.
Theroleplayedbyspecificsubjectsinthegenerationofdataflowsisthe mainreasonforthisconcentration.Governmentsandbigprivatecompanies (e.g.,largeretailers,telecommunicationcompanies)collecthugeamountsof datawhileperformingtheirdailyactivities.Thisbulkofinformationrepresentsastrategicandeconomicallyrelevantasset,asthemanagementoflarge databasesenablestheseentitiestoassumetheroleofgatekeeperswithregard totheinformationthatcanbeextractedfromthedatasets.Theyareableto keepinformationcompletelyclosedortolimitaccesstothedata,perhaps tospecificsubjectsonlyorwithregardtocircumscribedpartsoftheentire collection.
Notonlygovernmentsandbigprivatecompaniesacquirethispowerbut alsotheintermediariesininformationflows(e.g.,searchengines,Internet providers,databrokers,andmarketingcompanies),whichdonotgenerate informationbutplayakeyroleincirculatingit.
Therearealsodifferentcasesinwhichinformationisaccessibletothepublic,bothinrawandprocessedform(e.g.,opendatasets,onlineuser-generated contents).Thisonlyapparentlydiminishestheconcentrationofpowerover information,asaccesstoinformationisnotequivalenttoknowledge[51].
Alargeamountofdatacreateknowledgeifthedataholdershavethe adequateinterpretationtoolstoselectrelevantinformation,toreorganizeit, toplacethedatainasystematiccontext,andiftherearepeoplewiththe requiredskillstodefinethedesignoftheresearchandgiveaninterpretationto theresultsgeneratedbyBigDataanalytics[3,15,52,53].Withouttheseskills, dataonlyproduceconfusionandlessknowledgeintheend,withinformation interpretedinanincompleteorbiasedway.Forthesereasons,theavailability
∗ Thenoticedescribeshowthedataareprocessedandthedetailedpurposesofdata processing.
† SeeArticles2(h),7(a)and10,Directive95/46/EC.
ofdataisnotsufficientintheBigDatacontext[54,55].Itisalsonecessaryto havetheadequatehumanandcomputingresourcestomanageit.
Inthisscenario,controloverinformationdoesnotonlyregardlimited accessdata,butcanalsoconcernopendata[56,57],overwhichtheinformationintermediariescreateanaddedvaluebymeansoftheirinstruments ofanalysis.Giventhatonlyfewentitiesareabletoinvestheavilyinequipmentandresearch,thedynamicsdescribedearlierenhancetheconcentrationofpoweroverinformation,whichincreasesduetothenewexpansionof BigData.
Undermanyaspects,thisnewenvironmentresemblestheoriginsofdata processing,when,inthemainframeera,technologieswereheldbyafew entitiesanddataprocessingwastoocomplextobeunderstoodbydatasubjects.Nevertheless,thereareimportantdifferencesthatmayaffectthepossibleevolutionofthissituation,intermsofadiffusedanddemocraticaccessto information.
Thenewdatagatherersdonotbasetheirpositiononlyonexpensivehardwareandsoftware,whichmaybecomecheaperinthefuture,orisbasedonthe growingnumberofexpertsabletogiveaninterpretationtotheresultsofdata analytics.Thefundamentalelementofthispowerisrepresentedbythelarge databasestheyhave.Thesedatasilos,whichareconsideredthegoldmineof thetwenty-firstcentury,donothavefreeaccess,astheyrepresentthemain orthesideeffectoftheactivitiesconductedbytheirowners,duetotherole thattheyplayincreating,collecting,ormanaginginformation.
Forthisreason,intheBigDatacontext,itseemsquitedifficulttoimagine thesameprocessof democratization thathappenedwithregardtocomputer equipmentduringthe1980s[58].Theaccesstolargedatabasesisnotonly protectedbylegalrights,butitisalsostrictlyrelatedtothepeculiarpositions heldbydataholdersintheirmarketandtothepresenceofentrybarriers.
Anotheraspectthatcharacterizesthisnewformofconcentrationofcontroloverinformationisthenatureofthepurposesofdatacollection:data processingisnolongerfocusedonsingleusers(profiling),butitincreasedby scaleanditistryingtoinvestigateattitudesandbehaviorsoflargegroupsand communities,uptoentirecountries.Theconsequenceofthislarge-scaleapproachisthereturnofthefearsaboutsocialsurveillance,whichcharacterized themainframeera.
Againstthisbackground,theGDPRdoesnotchangethemainpillars ofthepreviousregulatorymodel.Therefore,personaldataarestillprimarilyprotectedbyindividualrights;the noticeandconsent modelremainsan importantlegalgroundfordataprocessing,andtheprinciplesofpurposelimitationanddataminimizationarereaffirmed.
Despitethistraditionalapproach,whichseemstobepartiallyinadequateintheBigDatacontext,theGDPRshowsapartialshiftofthe regulatoryfocusfromdatasubject’sself-determinationtoaccountabilityof thecontrollerandpersonsinvolvedindataprocessing.Inthissense,accountabilityrepresentsthecoreofthenewEUdataprotectionframework
andanimportantelementtotacklethepotentialnegativeimpactsofthe useofdataanalytics[59].
Morespecifically,accountabilityisbasedonthedataprotectionimpact assessment,theroleplayedbydataprotectionofficersand,whenrequiredby law,thepriorassessmentprocessconductedbydataprotectionauthorities.In thissense,comparedwiththepreviousDataProtectionDirective,theGDPR undoubtedlymovestowardarisk-basedapproach.
Nevertheless,thistransitionisstillincomplete.Elementsoftheprevious modelfocusedondatasubjectsthatcoexistwiththenewapproach,butwithoutacompleteredraftofthearchitecturedefinedinthe1990s,itseemstobe difficulttoaddressthesocialandtechnologicalchallengesofBigData.
Useofdataandrisk-analysis
Regardingriskmanagementindataprocessing,itisworthpointingout thatriskcanbeconsidered,inabroadsense,asanynegativeconsequence thatcanoccurwhenpersonaldataareprocessed,regardlessofthefactthat theseconsequencesmightproducedamageorprejudicetoindividualrights andfreedoms.
Inthissense,datasubjectsthatusesocialnetworksexposethemselvesto theriskofbeingprofiled[60],ofhavingtheirinformationsharedwiththird parties,ofbeingtrackedforcommercialpurposes,andsoon.Noneofthese consequencesareagainstthelaw,asthosearedetailedintermsandconditions andprivacypoliciesbyserviceprovidersandacceptedbyusers,onthebasis ofthe noticeandconsent model.
Inthesecases,itseemsthatthereisnorelevantriskforthesafeguard ofdatasubjects’rights,asindividualscanassesstheconsequencesofdata processingandhavefreelyexpressedtheirconsent.Nevertheless,legaland sociologicalstudieshaveclearlydemonstratedthatusersareusuallyunaware oftheconsequencesofprovidingtheirconsent,astheydonotreadlongand technicalnoticesorarenotabletocompletelyunderstandthesedescriptions andimaginetheirpracticalconsequences[61–65].Moreover,inmanycases, powerimbalanceandsociallock-indrasticallyreduceanyeffectivefreedom ofchoice.
Asaconsequenceoftheseconstraints,usersfrequentlyacceptsomeforms ofdataprocessingwithoutanypriorrisk/benefitanalysisandareunawareof theconsequences.Thisshowsthelimitsofthetraditional noticeandchoice paradigm[66,67],whicharemoreevidentinthecontextofBigDataanalytics, inwhichitisdifficulttodescribethe“specific”purposesofdataprocessing [Article6(1)(a)GDPR]atthemomentofdatacollection,duetothetransformativeuseofdatamadebydatacontrollers[68].∗
∗ Inthislight,itisalsodifficulttocomplywiththeprovisionsofArticle4oftheGDPR, whichqualifiesdatasubject’sconsentas“freelygiven,specificandinformed.”Accordingto theArticle29DataProtectionWorkingParty,“tobespecific,consentmustbeintelligible:it shouldreferclearlyandpreciselytothescopeandtheconsequencesofdataprocessing”[17].
Inthissense,withrespecttothebroadnotionofrisk-concerningdataprocessing,theGDPRmaintainstheimportantrolesplayedbyself-determination ofdatasubjectsandtransparency,recognizedbylawinthelastdecades. TheEuropeanlegislatorseemstobeunawareoftheweaknessesofthis approach,wheretheformaltransparencyoftermsandconditionscombinedwithusers’behavior[61]providedatacontrollerswiththe noticeand consent model,aneasywaytolawfullyexploitpersonaldatainanextensive manner.
Ontheotherhand,anarrowernotionofriskcanbeadopted,whichfocuses on“materialornonmaterialdamages”thatprejudicethe“rightsandfreedom ofnaturalpersons.”ThisnotionhasbeenadoptedintheGDPRtodefinethe risk-basedapproach(Recital75GDPR).Accordingtotheregulation,whena riskofprejudiceexistsandcannotbemitigatedorexcluded,dataprocessing becomesunlawful,despitethepresenceofanylegitimategrounds,suchasthe datasubject’sconsent.
Recitaln.75oftheGDPRprovidesalonglistofcasesinwhichdataprocessingisconsideredunlawful.Moreover,thisrecitaldoesnotlimitthesehypothesestothesecurityofdataprocessingbutalsotakesintoaccounttherisk ofdiscriminationand“anyothersignificanteconomicorsocialdisadvantage.”
Thisnotionofriskimpact,whichisechoedintheArticle35oftheGDPR, representsanimportantstepinthedirectionofanimpactassessmentofdata processing[69]thatisnolongerprimarilyfocusedondatasecurity(seeArticle 32GDPR)andevolvestowardamorerobustandbroaderPrivacy,Ethical,and SocialImpactAssessment(PESIA).∗ Moreover,theattentiontotheeconomic andsocialimplicationsofdatausesassumesrelevanceintheBigDatacontext,inwhichanalyticsareusedindecision-makingprocessesandmayhave negativeimpactsthataffectindividualsintermsofdiscriminationratherthan intermsofdatasecurity.†
Inlinewiththerisk-basedapproach,thenewprovisionsoftheGDPR reinforcetheaccountabilityofdatacontrollersthat,accordingtoArticle24,are liablewhentheydonot“implementappropriatetechnicalandorganizational measures”totackletherisksmentionedintheregulation(seealso Article83(4)GDPR).Thesemeasuresshouldbeimplementedfromthe earlieststageofdataprocessingdesign,embeddingthemintheprocessing, accordingtothedataprotectionbydesignapproach(Article25GDPR).
Inthelightoftheabove,regardingtransparency,rightstoaccess,anddata protectionauthorities,whicharethefoundingpillarsofdataprotectionregulation,andthefurtherelementofthedatasubject’sconsent,thenewregulation
∗ Seesections“Multiple-riskassessmentandcollectiveinterests”and“Theguidelines adoptedbytheCouncilofEuropeontheprotectionofindividualswithregardtothe processingofpersonaldatainaworldofBigData.”RegardingthePESIAmodel,seealso theH2020project“VIRT-EU:ValuesandethicsinInnovationforResponsibleTechnology inEurope.”http://www.virteuproject.eu/(accessedDecember21,2016).
† Seesection“Data-centeredapproachandsocio-ethicalimpacts.”
shedslightontheaccountabilityofdatacontrollers.Althoughaccountability principleswerealreadypresentinthefirstdataprotectionregulations,in whichthedutiesoftransparencyandtheroleplayedbydataprotectionauthoritiesincreaseddatacontrollers’accountability,intheDirective95/46/EC, therewasnotageneralprocessofrisk-assessment,withspecificconsequences intermsofaccountability.
Beforethenewregulation,therewereonlynationalprovisionsorbest practicesregardingtheprivacyimpactassessment[69],butnouniformriskbasedapproach.ThisgoalhasnowbeenreachedintheGDPRbymeansofa setofrulesthatconcerntheroleplayedbyriskanalysis,thedataprotection impactassessment,thepriorconsultationofdataprotectionauthorities,and thedataprotectionofficer(Articles35,36,and37GDPR).
Inmoredetail,therisk-basedmodeldefinedbytheGDPRisarticulatedin threedifferentlevelsofassessment.ThefirstisrequiredbyArticle24GDPR, andimplicitlybyArticle35(1).Thisisageneralassessmentof“theriskof varyinglikelihoodandseverityforrightsandfreedomsofnaturalpersons,” whichdefinesthelevelofthepotentialnegativeimpactofdataprocessing.
Whenthisfirstassessmentshowsthattheprocessing“islikelytoresultin ahighrisktotherightsandfreedomsofnaturalpersons”(Article35GDPR), thecontrollershouldcarryoutaformaldataprotectionimpactassessment. Moreover,thereisalistofcasesinwhichhighriskispresumed(Article35(3) GDPR).Thisisanopenlist,duetothefactthatdataprotectionauthoritiesmayaddfurthercases(Article35(4)GDPR),accordingtothemargin ofmaneuverrecognizedinseveralprovisionsbytheregulationtonational authoritiesorlegislators.
Nevertheless,theideaofalistofhigh-riskcases,aswellasofcasesexcluded fromtheimpactassessment(Article35(5)GDPR),raisedoubtsaboutthe feasibilityofthiscategorization.Inthissense,an exante generaldefinition ofthepresumedlevelofriskseemstobeinconflictwiththeideaofriskassessment,whichisnecessarilycontextbased.
Moreover,thecasesofhighriskaredescribedusingindefinitenotions, suchas“largescale”dataprocessing(Article35(3)(b)and(c)GDPR).Inthis regard,Recitaln.91maybeofhelptoclarifythemeaningofthisprovision,as itstatesthattheimpactassessment“shouldinparticularapplytolarge-scale processingoperationswhichaimtoprocessaconsiderableamountofpersonal dataatregional,nationalorsupranationallevelandwhichcouldaffectalarge numberofdatasubjects.”Nevertheless,therecitaldoesnotexplainwhen anamountofdataisdeemed“considerable”andwhy,inthedigitalglobal context,theamountofdatashouldrefertoterritorialdimensions(regional, national,orsupranational).
Finally,intheabsenceofanyscale,thegeneralnotionof highrisk remains quiteindefinite.Recitaln.77identifiesaseriesofbodiesandinstrumentsthat canprovideguidanceasregardsthe“identificationoftheriskrelatedtothe processing,theirassessmentintermsoforigin,nature,likelihoodandseverity,” but,atthemoment,theframeworkremainsuncertain.
ThesecriticismsseemtohavealimitedimpactonthefieldofBigData analytics,asthemajorityofapplicationsfallwithinthecaseslistedin Article35(3)GDPR,inwhichhighriskispresumed.Nevertheless,itisworth pointingoutthatanalyticscanbeusedincontextsinwhichtheevaluationof personalaspectsisnotnecessarily“systematicandextensive,”astheymay focusonlyonaspecificsubsetofattributesoronagivenclusterofpersons.
PursuanttoArticle35(3),theuseofBigDataanalyticsusuallyrequires apriordataprotectionimpactassessment.Thisprocedureisdefinedby Article35(7),inlinewiththetraditionalmodelofrisk-assessment,whichis primarilyapriorevaluationofthepotentialnegativeoutcomesofaprocess, product,oractivity,andaconsequentidentificationofthemeasuresthat shouldbeadoptedtoavoidor,atleast,mitigatetheidentifiedrisks.∗
Thisprocedurecanbedividedintothreedifferentstages:analysisof theprocess(Article35(7)(a)GDPR),risk-assessment(Article35(7)(b)and (c)GDPR),anddefinitionofthemeasuresenvisagedtoaddresstherisks (Article35(7)(d)GDPR).Itisworthpointingoutthatthestageconcerning therisk-assessmentincludestwodifferentkindsofevaluation:assessmentof the“necessityandproportionality”ofdataprocessing,andassessmentofthe “riskstotherightsandfreedomsofdatasubjects.”Thesetwoevaluationsare correlatedandconsequent,asdisproportionalorunnecessarydataprocessing cannotbeputinplaceand,inthiscase,thereisnotanyfurtherquestionabout theimpactonindividualrightsandfreedoms.Ontheotherhand,whenthe principlesofnecessityandproportionalityarerespected,furtherinvestigation isneededtoassessthespecificbalanceofintereststhattheuseofdataimplies.
AccordingtotheprinciplesandvaluesframedintheEuropeanChart ofFundamentalRightsoftheEuropeanUnion,thisbalanceofinterestsis notamererisk/benefitanalysis,butacomparisonbetweenintereststhat aredifferentandmayhaveadifferenthierarchicalorder.† Inthissense,the dataprotectionimpactassessmentisnotinlinewiththerisk-basedtheories [70]thatsuggesttheadoptionofarisk/benefitapproachinsteadofariskmitigationapproach.‡
∗ Accordingtothetraditionalparadigmofrisk-assessment,datacontrollersshouldbe abletodemonstratecompliancewiththeRegulationonthebasisoftheassessmentresults (Article35(7)(d)GDPR)andshouldperiodicallyreviewtheseresults,duetothepossibility ofachangeinthenatureandseverityoftherisksoverthetime(Article35(11)GDPR).
† SeeEuropeanCourtofJustice,May13,2014,Case131/12, GoogleSpainSL, Google Inc.vAgenciaEspa˜noladeProtecci´ondeDatos(AEPD),MarioCostejaGonz´alez http://curia.europa.eu/juris/document/document.jsf?text=&docid=152065&pageIndex=0 &doclang=EN&mode=lst&dir=&occ=first&part=1&cid=980962(accessedJune16,2016).
‡ Accordingtotherisk/benefitapproach,theassessmentshouldbebasedonthecomparisonbetweentheamountofbenefitsandthesumofallrisks,withoutanydistinction regardingthenatureofrisksandbenefits.Inthissense,forinstance,economicbenefitsmay prevailoverindividualrights.Ontheotherhand,theriskmitigationapproachassumes thatsomeinterests(e.g.,fundamentalrights)areprevailingandcannotbecomparedwith otherintereststhathavealowerrelevance.Asaconsequence,theriskmitigationapproach focusesonthepotentialprejudiceforfundamentalrightsandsuggestsadequatemeasures toreducethisriskor,wherefeasible,toexcludeit.
Whendataprotectionimpactassessment“indicatesthattheprocessing wouldresultinahighriskintheabsenceofmeasurestakenbythecontrollertomitigatetherisk,”datacontrollersmustconsultthesupervisory authoritypriortothestartofprocessingactivities(Article36(1)GDPR). AccordingtoRecitaln.84oftheGDPR,theabsenceofmeasurestomitigate theriskisevaluatedtakingintoaccountthe“availabletechnologyandcosts ofimplementation.”
Itisworthpointingoutthatthereferencetothecostsandtheavailable technology,alsopresentintheprovisionsconcerningsecurityrisk(Recital n.83andArticle32(1)GDPR)anddataprotectionbydesign(Article25(1) GDPR),representsanimportantopportunitytoputtheprincipleofproportionalityintopracticeinthecontextofriskmitigation.Therefore,these provisionsreducetheriskofanexcessiveburdenfordatacontrollersdueto theimplementationoftherisk-assessmentmodel.
Whenadataprotectionimpactassessmentindicatesthatprocessingwould resultinahighriskintheabsenceofmeasurestakenbythecontrollerto mitigatetherisk,datacontrollersshouldconsultthesupervisoryauthority priortothestartofprocessingactivities(Recitaln.94GDPR).∗
AccordingtoArticle36(2)GDPR,whenthesupervisoryauthorityisof theopinionthattheintendedprocessingwouldinfringetheregulation,the authority“shall[...]providewrittenadvicetothecontrollerand,whereapplicabletotheprocessor,mayuseanyofitspowersreferredtoinArticle58.” GiventhepowersgiventosupervisoryauthoritiesbyArticle58,thismeans thattherearetwooptionsasfollows:(1)Theassessmentisnotsatisfactory, andthedatacontrollerhasnotadequatelyidentifiedormitigatedtherisk; (2)theassessmenthasbeenconductedinacorrectmanner,butthereare nomeasuresavailabletomitigatetherisk.Inthefirstcase,thesupervisory authorityordersthecontrollerorprocessor“tobringprocessingoperations intocompliancewiththeprovisionsofthisRegulation,whereappropriate,in aspecifiedmanner”(Article58(2)(d)GDPR),whereas,inthesecondcase, theauthorityimposes“atemporaryordefinitivelimitationincludingaban onprocessing”(Article58(2)(f)GDPR).
Finally,minoraspectsconcerningtherisk-basedapproachregardtherole playedbythedataprotectionofficer,whosemaintasksaretoprovideadvice tothecontrollerortheprocessoroftheirobligations(includedthedataprotectionimpactassessment),andtomonitorcompliancewithlegalprovisions concerningdataprotectionandwiththeprivacypoliciesofthecontrolleror processor(Article39(1)GDPR).Intheperformanceofthesetasks,thedata protectionofficermust“havedueregardtotheriskassociatedwithprocessing operations,takingintoaccountthenature,scope,context,andpurposesof
∗ Themodelofpriorconsultationisbuiltontheconceptofpriorchecking,whichwas alreadypresentinArticle20oftheDirective95/46/EC.
processing”(Article39(2)GDPR).Therefore,therisk-assessmentrepresents oneofthemaincriteriathatshoulddrivetheactionofthedataprotection officer.
Thenewprovisionsaboutrisk-assessmentrepresentanimportantevolutioninthedirectionofarisk-basedapproachindataprotectionand,inthis sense,mayofferanadequatesolutiontothepotentialnegativeoutcomesof theuseofBigDataanalytics.Themainlimitoftheseprovisionsliesinthe linktothepurposesofdataprocessing.∗
Althoughtheassessmentshouldnecessarilyberelatedtotheuseofdata foraspecificpurpose,thereisaproblemduetothefactthat,accordingto Article5(1)(b)GDPR,dataprocessingpurposesshouldbe“specific,explicit, andlegitimate”anddefinedatthemomentofdatacollection,whichcontrast withthetransformativeuseofdatamadebyprivateandpublicbodiesby meansofBigDataanalytics.
Forthesereasons,abetterdesignoftheimpactassessmentshouldnotfocus ontheinitialpurposeofdatacollection,butoneachspecificdatausethat isputinplacebythedatacontrollerafterdatacollection.Inthisregard,it shouldbenotedthat,atthemoment,thisresultisachievedbydatacontrollers circumventingtheprovisionsonpurposelimitation.Theycollectpersonaldata onthebasisofbroadseriesofdifferentpurposesandthen,iftheyhavealready adoptedproceduresofimpactassessment,evaluatecase-by-casethepotential impactondataprotection,withregardtoeachdifferentuseofinformation foragivenpurpose.
Againstthisbackground,adifferentperspectivecanbeadopted,which expresslyacceptstheideathatdataarecollectedformultiplepurposes,defined onlybroadlyatthebeginningofdataprocessing.Thismodelfocusesonthe differentspecificusesofcollectedinformationandthepriorassessmentofthe potentialrisksofeachuse.
Thiskindofapproach,ifadoptedbythelegislator,willbemoreefficient andconsistentwiththetransformativeuseofdatamadebycompaniesinthe BigDatacontext,aswellaswiththelevelofself-determinationofthedata subjects[66,71].Inthissense,amoreextensiveuseofthelegitimateinterest aslegalgrounds[24]maycompletethismodel.Companiesmayenlistusers indataprocessingwithoutanypriorconsent,providedtheygivenoticeof theresultsoftheassessment,whichshouldbesupervisedbydataprotection authorities(licensingmodel),andprovideanopt-outoption[66].
Itmightbenotedthatthesuggestedapproachunderminesthechancesfor userstonegotiatetheirconsent,butthestrengthofthisobjectionisreduced bytheexistinglimitstoself-determinationdescribedabove.Inthemajority
∗ SeeArticle35(1)GDPR(“Whereatypeofprocessinginparticularusingnewtechnologies,andtakingintoaccountthenature,scope,contextandpurposesoftheprocessing,is likelytoresultinahighrisktotherightsandfreedomsofnaturalpersons”)and35(7)(b) (“[Theassessmentshallcontainatleast]anassessmentofthenecessityandproportionality oftheprocessingoperationsinrelationtothepurposes”).
ofthecases,thenegotiationisreducedtothealternative takeitorleaveit Apriorassessmentconductedunderthesupervisionofindependentauthorities,theuseoflegitimateinterestaslegalground,andtheadoptionofan opt-outmodelseemtooffermoreguaranteestousersthananapparent,but inconsistent,self-determinationbasedon noticeandconsent andontheopt-in model.
Ontheotherhand,remainingfocusedontheexistinglegalframework definedbytheRegulation2016/679,adifferentoption[71]maybetolimit BigDatausestostatisticalpurposes,whichbenefitfromanexplicitlypermittedreuseofdata(Articles5(1)and89,GDPR).Nevertheless,inthiscase, usinganalyticsfordecision-makingpurposesdirectlyaffectingaparticular individualwouldbeoutsidethefieldofstatisticalpurposesandalsoviolate therestrictionsonautomatedindividualdecisionmaking,includingprofiling. Inthissense,theGDPR“canbeseenasasteppingstone,pointingtoward theneedtoevolvedataprotectionbeyondtheoldparadigm,yetnotfully committedtodoingso”[71].
ThemodelofdatamanagementdefinedbythenewRegulationdoesnot completelyaddressthenewchallengesofuseofBigDataanalyticsindata processing[24,71]:thenewprovisionsdonotprovideaneffectivetransparency ofdataprocessing(obscurenotices,impactassessmentnotpubliclyavailable), butonlyahigherlevelofaccountability.
Moreover,therisk-mitigationapproachadoptedbytheRegulationseems stilltobefarfromtheideaofamultipleandparticipativerisk-assessment. AlthoughRecitaln.75recognizestheriskofdiscriminationand“anyother significanteconomicorsocialdisadvantage,”theprovisionsoftheRegulation donotofferanadequateframeworkfortheassessmentofthiskindofnegative outcome.
WithregardtotheuseofBigDataanalyticsindecision-makingprocesses, importantquestionsariseabouttheethicalandsocialvaluesthatshouldbe takenintoaccount,aswellastherolethatthedifferentsocialstakeholderscan playinassessingtheimpactofdatauses.∗ Inconclusion,theEuropeanUnion seemstobeinsecureinmovingitsstepsawayfromthetraditionalmodelof dataprotection,whereasotherinternationalbodiesaretryingtoofferamore courageousanswertothechallengesofthedataage.
Inthissense,thenewguidelinesonBigDataoftheCouncilofEurope seemtobeawareofthelimitsofthetraditionalprinciplesgoverningdata protectionandopentoabroaderrisk-assessment,whichtakesintoaccount thesocialandethicalimpactsofdatausesandrecognizesthebenefitsofa participatorymodelbasedonthemultistakeholderapproach.†
∗ Seesection“TheguidelinesadoptedbytheCouncilofEuropeontheprotectionof individualswithregardtotheprocessingofpersonaldatainaworldofBigData.”
† Seesection“Multiple-riskassessmentandcollectiveinterests.”
Useofdatafordecision-makingpurposes:Fromindividual tocollectivedimensionofdataprocessing
ThenewscaleofdataprocessingofBigDataapplicationsandtheuseof analyticsindecision-makingprocessesposenewquestionsaboutdataprotection.AsBigDatamakeitpossibletocollectandanalyzelargeamountsof information,dataprocessingisnolongerfocusedonindividualusers,andthis shedslightonthecollectivedimensionoftheuseofdata.
IntheBigDataenvironment,generalstrategiesareadoptedonalarge scaleandonthebasisofrepresentationsofsocietygeneratedbyalgorithms, whichpredictfuturecollectivebehavior[3,25,55,64].Thesestrategiesarethen appliedtospecificindividuals,giventhefactthattheyarepartofoneormore groupsgeneratedbyanalytics[3,56,72].
Theuseofanalyticsandtheadoptionofdecisionsbasedongroupbehavior ratherthanonindividualsarenotlimitedtocommercialandmarketcontexts. Theyalsoaffectotherimportantfields,suchassecurityandsocialpolicies, whereadifferentbalanceofinterestshouldbeadopted,giventheimportance ofpublicinterestissues.∗ Oneexampleofthisisprovidedbypredictivepolicing solutionssuchas PredPol [73–77].
This categorical approachcharacterizingtheuseofanalyticsleadspolicymakerstoadoptcommonsolutionsforindividualsbelongingtothesame clustergeneratedbyanalytics.Thesedecisionalprocessesdonotconsiderindividuals perse,butasapartofagroupofpeoplecharacterizedbysome commonqualitativefactors.
Inthissense,theuseofpersonalinformationandBigDataanalyticsto supportdecisionsexceedstheboundariesoftheindividualdimensionand assumesacollectivedimension[78],withpotentialharmfulconsequencesfor somegroups[79,80].Inthissense,prejudicecanresultnotonlyfromthewellknownprivacy-relatedrisks(e.g.,illegitimateuseofpersonalinformation,data security)butalsofromdiscriminatoryandinvasiveformsofdataprocessing [15,81,82].
Thedichotomybetweenindividualsandgroupsisnotnew,andithas alreadybeenanalyzedwithregardtothelegalaspectsofpersonalinformation. Nonetheless,therighttoprivacyandtherighttotheprotectionofpersonal datahavebeenlargelysafeguardedasindividualrights,despitethesocial dimensionoftheirrationale.
Thefocusonthemodelofindividualrightsisprobablythemainreason forthefewcontributionsbyprivacyscholarsonthecollectivedimensionof privacyanddataprotection.Hitherto,onlyfewauthorshaveinvestigatedthe notionofgroupprivacy.Theyhaverepresentedthisformofprivacyastheprivacyofthefactsandideasexpressedbythemembersofagroupinthegroup environmentorintermsofprotectionofinformationaboutagroup[37,83,84].
∗ Seealsosection“Dataprediction:socialcontrolandsocialsurveillance.”
Ontheotherhand,collectivedataprotectiondoesnotnecessarilyconcernfactsorinformationreferringtoaspecificperson,aswithindividual privacyanddataprotection.Nordoesitconcernclustersofindividualsthat canbeconsideredasgroupsinthesociologicalsenseoftheterm.Inaddition, collectiverightsarenotnecessarilyalarge-scalerepresentationofindividualrightsandrelatedissues[85].Finally,collectivedataprotectionconcerns non-aggregativecollectiveinterests[86],whicharenotthemeresumofmany individualinterests.∗
Theimportanceofthiscollectivedimension[78]dependsonthefactthat theapproachtoclassificationbymodernalgorithmsdoesnotmerelyfocuson individuals,butongroupsorclustersofpeoplewithcommoncharacteristics (e.g.,customerhabits,lifestyle,onlineandofflinebehavior).Datagatherersare mainlyinterestedinstudyinggroups’behaviorandpredictingthisbehavior, ratherthaninprofilingsingleusers.Data-drivendecisionsconcernclusters ofindividualsandonlyindirectlyaffectthemembersoftheseclusters.One exampleofthisispricediscriminationbasedonage,habits,orwealth.
Themostimportantconcerninthiscontextistheprotectionofgroups frompotentialharmduetoinvasiveanddiscriminatorydataprocessing.In thissense,thecollectivedimensionofdataprocessingismainlyfocusedon theuseofinformation[66,70],ratherthanonsecrecy[83,84]anddataquality. Regardingtheriskofdiscrimination,thissectiondoesnotfocusonthe unfairpracticescharacterizedbyintentionaldiscriminatorypurposes,which aregenerallyforbiddenandsanctionedbylaw[87,88],† butontheinvoluntary formsofdiscriminationincasesinwhichBigDataanalyticsprovidebiased representationsofsociety[89,90].
Forexample,in2013,astudyexaminedtheadvertisingprovidedbyGoogle AdSenseandfoundstatisticallysignificantracialdiscriminationinadvertisementdelivery[91,92].Similarly,KateCrawfordhaspointedoutcertain algorithmicillusions [93,94]anddescribedthecaseoftheCityofBostonand itsStreetBumpsmartphoneapptopassivelydetectpotholes[95].‡
AnotherexampleistheProgressivecase,inwhichaninsurancecompany obligeddriverstoinstallasmallmonitoringdeviceintheircarstoreceivethe
∗ ContraVedder[81],whoclaimsthatthenotionofcollectiveprivacy“remindsofcollectiverights,”butsubjectsofcollectiverightsaregroupsorcommunities.Conversely,the groupsgeneratedbygroupprofilingarenotcommunitiesofindividualssharingsimilar characteristicsandstructuredororganizedinsomeway.Forthisreason,Vedderusesthe differentdefinitionof“categorialprivacy.”
† SeeArticle14oftheConventionfortheProtectionofHumanRightsandFundamental Freedoms;Article21oftheCharterofFundamentalRightsoftheEuropeanUnion;Article 19oftheTreatyontheFunctioningoftheEuropeanUnion;Directive2000/43/EC;Directive 2000/78/EC.
‡ Inthiscase,theapplicationhadasignalproblem,duetothebiasgeneratedbythe lowpenetrationofsmartphonesamonglowerincomeandolderresidents.WhiletheBoston administrationtookthisbiasintoaccountandsolvedtheproblem,less-enlightenedpublicofficialsmightunderestimatesuchconsiderationsandmakepotentiallydiscriminatory decisions.
Another random document with no related content on Scribd:
»Mene heti sisään äitimuorin luo!» sanoi hän. »Etkö ymmärrä, että heidän täytyy laittaa vuode kuntoon, jotta meillä olisi minne laskea hänet, kun palaamme takaisin?»
Jannen täytyi siis mennä sisään Fallan emännän puheille, ja vaikka hän koettikin kiirehtiä, niin aikansa se sittenkin kysyi, kun oli kerrottava koko tapaus juurta jaksain.
Kun Janne palasi takaisin pihalle, kuuli hän Larsin meluavan ja kiroilevan tallissa. Lars ei kohdellut hyvin eläimiä. Hevoset potkivat, niin pian kun hän tuli lähellekin. Nytkään hän ei ollut kyennyt saamaan ainoatakaan hevosta ulos pilttuusta koko sinä aikana, jolloin Janne oli puhunut Fallan emännän kanssa.
Siitä ei koituisi mitään hyvää, jos Janne yrittäisi auttaa häntä, sen hän ennestään tiesi, siksipä hän läksi toista asiaa toimittamaan, renkipoikaa hakemaan. Omituista, ettei Lars ollut käskenyt häntä pikemmin antamaan sanaa Börjelle, joka oli riihessä puimassa, vaan lähetti hänet poikaa kutsumaan, joka raivasi nuorta metsää koivikossa hyvän matkan päässä talosta.
Heikko ääni kuusen alta kaikui Jannen korvissa hänen toimittaessaan näitä turhia asioita. Se ei kuulunut enää niin käskevältä, vaan se rukoili ja pyysi häntä pitämään kiirettä. »Kyllä tulen, kyllä tulen», kuiskasi Janne vastaan, mutta samalla hänet valtasi samallainen tunne kuin unissa, painajaisen painostaessa, jolloin ponnistaa kaikki voimansa päästäkseen eteenpäin, mutta ei sittenkään voi hievahtaa paikaltaan.
Nyt Lars oli saanut hevosen valjaihin, mutta sitten naisväki tuli sanomaan, että hänen pitäisi ottaa olkia ja peitteitä mukaan, ja
sehän oli kylläkin hyvä, vaikka tuottihan sekin viivytystä, ennenkuin kaikki oli järjestyksessä.
Vihdoin he läksivät liikkeelle, Lars, Janne ja renkipoika, mutta he eivät päässeet metsänreunaa pitemmälle, ennenkuin Lars pysähdytti hevosen.
»Sitä joutuu aivan pyörälle päästään saadessaan tällaisia uutisia», sanoi hän. »Nyt vasta muistuu mieleeni, että Börje on riihellä.» »Niin», sanoi Janne. »Hän olisi ollut hyvä matkassa. Hän on kahta vertaa voimakkaampi kuin kukaan meistä.»
Silloin Lars käski renkipojan juosta hakemaan Börjeä, ja taaskin oli odotettava.
Sillä välin kuin Janne istui toimettomana reessä, tuntui hänestä, ikäänkuin hänen eteensä olisi avautunut suuri, tyhjä, jäisen kylmä kuilu, johon oli kamala katsoa. Mutta samalla se ei ollut mikään kuilukaan, vaan ainoastaan varma tunne, että he saapuisivat liian myöhään.
Börje ja poika tulivat vihdoin juoksujalkaa ja henki kurkussa, ja nyt he ajoivat metsään.
Mutta matka ei sujunut nopeasti. Lars oli valjastanut vanhan, jäykkäjalkaisen Ruskon reen eteen. Totta tosiaan, niinkuin hän itsekin sanoi, hän mahtoi olla aivan pyörällä päästään.
Hetken kuluttua osoittautui jälleen, että hän oli aivan sekaisin. Kesken kaikkea hän tahtoi ajaa väärää tietä. »Ei, jos me ajamme tuota tietä, niin me joudumme suoraan Snipavaaralle», sanoi Janne, »ja meidänhän on määrä päästä Lobyn tuolla puolen olevaan
metsään.» — »Niin, kyllä minä sen tiedän», sanoi Lars, »mutta kauempana on toinenkin oikotie, jota on parempi ajaa.» — »Mikähän oikotie se mahtaa olla?» kysyi Janne. »Sitä en koskaan ole nähnyt.» — »Maltahan vaan, niin saat nähdä.»
Lars alkoi pyrkiä suoraan vaaraa ylös. Mutta Börjekin piti Jannen puolta, ja silloin Larsin täytyi antaa perään. Joka tapauksessa oli kulunut aikaa hukkaan heidän kiistellessään, ja Janne tunsi, miten tuo musta tyhjyys levisi koko hänen ruumiiseensa. Käsivarret ja kädet tulivat niin ontoiksi ja kohmettuneiksi, että tuskin saattoi niitä liikutella. »Samapa se», tuumi hän. »Me tulemme kuitenkin liian myöhään. Fallan Erik ei kaipaa enää meidän apuamme, kun pääsemme perille.»
Vanha hevonen ponnisteli eteenpäin metsätiellä niin hyvin kuin saattoi, mutta sillä ei ollut kylliksi voimia sellaista matkaa varten. Se oli huonosti kengitetty ja kompastui kerta toisensa jälkeen, ja ylämäissä täytyi miesten nousta pois reestä ja astua jalan. Poikettuaan raivaamattomaan metsään oli hevosesta melkein enemmän haittaa kuin hyötyä.
Vihdoin he kuitenkin tulivat perille ja huomasivat, ettei Fallan Erik ollutkaan kovin pahoin vioittunut. Ei mikään jäsen ollut murskautunut eikä taittunut. Oksa oli repäissyt toista reittä, ja siinä oli kyllä vaikea haava. Mutta ei sekään ollut kuolemaksi.
Seuraavana aamuna, kun Janne tuli työhön, sai hän kuulla, että Erikillä oli korkea kuume ja kovia tuskia.
Hän oli vilustunut maatessaan maassa niin kauan aikaa. Tauti kääntyi keuhkokuumeeksi, ja kahden viikon kuluttua hän oli kuollut.
